The projected growth of the older adult population and the greater emphasis on early stage identification of Alzheimer's disease and similar dementias are placing heavy demands on the health care system, including the more than 15 million family members who provide some type of ongoing informal care to persons with dementia. A considerable body of research documents the primary and secondary stressors of dementia caregiving, and providing evidence-based services to ease the burden and stress associated with caregiving continues to be a major public health priority. Because family caregivers are an incredibly diverse group who differ along many demographic and situational variables, careful and sophisticated analyses are needed to better understand the mechanisms by which existing intervention trials have achieved their effects among diverse subgroups. In this proposal, we describe a series of secondary analyses to be conducted on the publically-available data from the second Resources for Enhancing Alzheimer's Caregivers Health (REACH II) project. REACH II was unique among the existing randomized trials of caregiver interventions in its enrollment of a geographically and racially diverse sample, but only a few publications to date have examined intervention outcome effects, and no published analyses have examined the mediators of intervention impact or tested whether such mechanisms differ by race or by other important demographic subgroups (e.g., gender, spouse vs. non-spouse caregivers). We propose a coordinated series of analyses that will address four inter-related aims. In particular, we propose 1) to conduct item-level psychometric analyses of the common instruments used to assess REACH II outcomes, including differential item functioning (DIF) analyses, to determine if some findings across demographic subgroups might be partially contaminated by measurement biases; 2) to perform mediation analyses of the potential mechanisms (improved social support, reduced stress appraisals) by which the intervention may have achieved primary outcome benefits (reduced caregiver depression and burden); 3) to conduct multiple group structural equation modeling analyses that will determine if the mediation effects from aim 2 and previously published moderator effects vary as a function of race or ethnicity; and 4) to examine the predictors of compliance with the REACH II intervention and conduct complier-average causal effect (CACE) analyses of the effect of the intervention among the subset of participants who sufficiently complied with it. These analyses will substantially inform the potential population impact of the REACH II intervention and be useful for guiding additional translational work for particular dementia caregiving subgroups.